I wanted to get your thoughts on how we can improve using a solver.

I think a lot of players basically look at solutions and backwards rationalize some weak heuristics based on their own misinformed biases and poor inputs. How would you go about studying some spot in a more rigorous fashion? How would you apply the scientific method to studying GTO solutons?

Step 1: Close the solver.
Step 2: Learn to make high EV plays. (Math, ranging, combinatorics, exploits, Hand vs range, future street projection...)
Step 3: Beat midstakes.
Step 4: Open the solver again.

I understand your point is to learn fundamentals first. But I think it misses the point of this post.

What percentage of the population do you think can beat midstakes (on a mainstream site like Stars or ACR) without studying modern GTO solutions?

U Gotta get a more aktiv learning process imo try and guess the solvers output before u see it and then see it the results match ur intuition.

Then it they don’t match u gotta figure out why- abs it usually has to do with some combination of:
Stack sizes
Range morphology
Vulnerability of strong hands across run outs
Phatt Equity advantage
Nut equity advantage - blockers
Lemon density
Apple density
Available bet and raise sizes
Etc

Then u can tweak inputs and see if outputs tweak in the expected way.

Now ur starting to get somewhere when u understand strategic incentives.

It’s all a fkn rabbit hole but I hope there’s a carrot at the bottom GL

Yeah I was being a bit hyperbolic, but I stand behind the fact that the utility of a solver is very niche and specific, and before playing highstakes, abusing and misapplying it is going to hurt your results while not touching a solver at all is not. For instance, the highest winrate reg I personally know that plays up to 2knl, doesn't use solvers at all, and I don't think he is an isolated case. I've given lessons to micros and small stakes players and one of the most common player types is the GTO maximalist that is struggling to beat the softest games online.

My honest recommendation for midstakes players and below is to not use a solver until they've mastered exploiting fish and weak regs (by far most regs). Not just because this is how most of the money is made in poker while the money from reg battling top regs is pretty inconsecuential, but because reading solver outputs with an exploitative mindset is going to shed light on so many things and make GTO so much easier to understand and this is what will help you apply it in high EV ways.

I think optimal studying is very, very different for every person. However being able to use logic to think through all aspects of a hand instead of just saying "oh I have a lot of better hands in my range here so I don't need to defend this" is more important than knowing GTO solutions.

I agree that a lot of winners don't use solvers, but I think basically 100% of midstakes+ crushers implement strategy heavily influenced by solvers, I mean at the very least I imagine he has probably looked at a solved preflop chart. Also our collective understanding of what betsizings/bet frequencies to play on various streets is heavily influenced by solvers. I don't remember the exact details but back when RedBaron was considered the GOAT, one of the other crushers in his game played a heavy limping style, something like 38/4 VPIP/RFI preflop. I don't think we will EVER see that again. On that note, personally I think there is so much secondhand information (much of which is free) out there that 99.9%+ of poker players and also 90%+ of poker professionals playing midstakes online would benefit more from studying through that rather than running all the solves themselves.

I mentioned this in my other GTO thread but one of the most common mistakes I see is someone running a solve for a specific spot they were put in to know if they made the right action. This seems pointless, I mean a channel like Finding Equilibrium might not have the exact flop texture and preflop position you want to study for whatever hand you're reviewing, but if you just understand the concepts presented then that is way more important than knowing "the right answer" (kind of reminds me of math class lol).

What has been most useful for me in solvers is finding and understanding concepts. For one example I looked at a multiway PLO spot on Q66r recently in a 3-bet pot. It has been conventional wisdom for years that on AAQ7 is "a better hand" than AAJ7 on Q66r, like why not right? But due to solvers now I understand that AAJ7 is actually a better hand to call a bet with because my opponent's betting range is polarized between Qxxx bluffs and 6xxx value and RARELY QQxx value, so I block almost all of his bluffs and relatively little of his value (and indeed, many players in real life do play this strategy because it is pretty intuitive from villain's perspective). It seems so much like common sense, but even just a few years ago many elite PLO players just wouldn't have thought of that. But here is the thing, you could have arrived at that conclusion using just logic and not a solver at all, but not at lot of people's brains work that way and I most certainly never discovered this seemingly simple concept in thousands of hours of studying PLO. I kind of feel like even if I was led directly to the water right before my eyes by someone asking me "is AAQ7 or AAJ7 on Q66r better?", I would have just been like "wtf obviously AAQ7 is better" and not thought more about it.

If you want to get a breakdown of what GTO looks like broadly, run sims across the 184 flop subset starting with the most common configurations, like BTN 2bets and BB defends, or CO defends to BTN 3b. You can run a tree that mimics the sizings used in your games. You can run some with as many sizings/options as your RAM allows, or use uncommon sizing options to see if they yield more EV than what you're used to.

Most solvers have built in filters/visualizers for processing that data and they also allow you to export the info into a spread-sheet. You can use Excel to work the data further. (Put "poker solver excel" into youtube and you'll get an idea. It helps me filter results faster and categorize by things like high-mid-low flops or mid-mid-low, etc.) At the very least, you'll get an idea of what flops are advantageous for whom, how often each flop gets c-bet, defended, re-raised etc... When you create a tree with multiple sizing options, you'll get an idea of what sizes are better on what textures. The more bet/raise sizing options the tree has, the closer it is to approximating the true Nash eq. You can use the solver to simplify strategies when you input a model with fewer sizing options that yields a similar EV to the most complex tree.

To get a better idea of the logic behind the numbers, it seems like most good players consider things like equity and nut advantage and the parts of your opponents range a specific holding may block. You can run toy-games to test your assumptions. For example, let one player have all TT+/AK combos and give the other player all combos of 55-99 and all suited connectors from 54s to T9s. Then test some flops that are good for one player, than the other, then maybe something that's in between. These extremes tend to make it clearer as to how to play when your range has favorable/unfavorable board textures. Also, test different SPRs. Stack depth obviously impacts your bet sizes and frequencies.

When focusing on a specific flop, maybe you can ask yourself what hands you would c-bet or x/r or whatever. Write it down, then compare to the solver output. Consider what you'd do on specific turns. Write it down, compare it to the solver. See how it constructs it's range and how it buckets hands into triple barrels, or what bluffs it had left on a certain river, etc.

If you play online, you should have a healthy database on your playerpool. You can start running sims to exploit it. Start out with their pf frequencies/ranges and input those to start with. Then work in some nodelocks to match their post-flop stats. See how your strategy changes when theirs's is locked in. Remember, a solver is set to exploit. The GTO strat it shows is the result of both players exploiting the other as much as possible until they really can't... or at least until they reach an EV difference threshold we set &or it's limited to. Anyways, run a player-pool report, filter for the flop texture you're studying and see if the pool is c-betting too much/too little, defending too much/too little. You might be able to see the ranges they're doing it with as well pending on your software. That just helps you create better nodelocks. You could also set nodelocks based on a certain type of player. For example, see how to counter people who x/r too much or float too wide by locking those frequencies for your opponent.

Basically, you never want to just run one sim and be done with it. Test out different bet/raise sizes on all streets. Try to make as simple of a game tree as you can that doesn't drop your EV too much from the "baseline" of as many possible sizes as your computing power allows. See how that strategy might change if the pf range of your opponent is different (tighter & wider). Test out some exploits by nodelocking. What are good responses to people more passive or more aggro than the solver's solution for them? Maybe nodelock yourself to test out strategies that are easier to implement...

Just some thoughts/ramblings.

edit window timed out for my post above, but so much of this is also about being needing a solver to be good is using the context of poker online. but I'd also like to add this quote which I agree with for live poker.

"I think a truly great live player can play nearly perfect because everyone kinda sucks, with perfect being defined in this case as making the correct decision against your opponent's exact hand."- gman06 post solvers, easily one of the biggest live poker winners still today.

He's got 43 VPIP short handed lul
But yeah him and his crew are guys that have been working on the math behind decisions in poker way before solvers existed and they probably understand GTO better than most people without ever looking at a solver

Here is one exercise I do from time to time.
You choose a certain flop and some line say BBvBTN 25Jr BTN bets1 / 3 and BB XC. After that, on paper, I write (what i think are) the top 5 highest EV turns for BU and 5 lowest EV and in rough outline what the strategies will look like. Then I compare it with the output solver and take a note if I missed a lot somewhere.

You can also make a more accurate version with Pokerranger. Where in the program for that spot you save for example the flop(what you think is) cbet range for BTN and then BB XR and XC and compare it with the solver.

This way you avoid the problem of rationalization after seeing the solver results.

I totally agree with aner0 that a solver is not the best learning tool for most people. Most players do not understand the basic math related to GTO play. People focus on things that bring very little EV. There are discussions on the micro stakes forum 2-3 pages long about which is the best flop sizing for one specific flop, and the difference in EV between two sisings is less than 0.5%.

We have a lot of better hands in our range there than KQ? Do you think a solver would x/f the river?

I've currently never used a solver directly, but I've been thinking about it to clean up my ranges in certain spots. I understand the idea to not using them until you get to higher stakes were your ranges need to be better ... but it also seems like you'll be practicing/re-enforcing bad ranges in spots the longer you go without looking at the "correct" answer.

I do not know the answer but from what I've seen I would be EXTREMELY surprised if it puts in 130bb 4-ways on this runout with 2nd pair no spades, yeah.

I mean, I can well believe that the robots don't bet as much as you did on flop/turn ... when I see solver lines the sizing is often the most confusing part.


But given that we did it and get to the river with KQ no clubs seems like a lol x/c. Only flop was 4 ways, does Ax or backdoor spades make up that much of a GTO flop calling range?

I understand that humans are probably going to way under/over bluff this river, I just assumed that GTO would do it enough that KQ no clubs was a call anyway. Hmm.

I feel like this is just something I know intuitively. Like of course if you block hands V might bluff with you're in worse shape facing his bet. I guess other spots will be less obvious though. What I'm trying to find out rn is if solver work would be beneficial for me, even though I'm a primarily live 2/5 player

Feels like we're saying the same thing over and over again. No one is going to know for sure what the solver equilibrium is unless someone has the 4 way sim. However, the closest thing I can attempt to model in a HU pot would even mix calls/folds with AQ (top two pair), so I suspect KQ is way too light.

ty!

The solver comes to conclusions based on value:bluff ratios, pretty simple except OTF + OTT... and factor that when you can use many different bet sizing lines across different streets given the increasing stack depth.

The spreadsheet stuff is sort of unnecessary imo unless you plan on playing HU or in a tough pool that will deviate very little from the nash strat. If they are deviating a lot it seems like a waste of time to study that just input nodelocks and pretty much use those intuition skills in real time